How do you measure (and govern) AI coding assistants across your team today?

We are launching MainLayer on Tuesday: one console for Copilot, Cursor, Claude Code and six other tools, showing adoption, cost, which merged PRs were AI-assisted, and blocking secrets on the developer machine before they reach a prompt. Metadata only, never code.

Before launch I would love to hear from engineering leaders and security folks: what do you use today, and what is the one question about AI tool usage you still cannot answer? Honest "we don't measure it" answers are the most useful. I will share what we learned building attribution without reading diffs in the comments.